Rotational allomerism and divergent evolution of domains in

Sep 1, 1975 - Allen B Edmundson , Christina R DeWitt , Benjamin Z Goldsteen , Paul A ..... Martin Giedlin , Nicholas R. J. Gascoigne , Christopher Goo...
1 downloads 0 Views 1MB Size
ROTATIONAL ALLOMERISM IN IMMUNOGLOBULIN LIGHT CHAINS

Rotational Allomerism and Divergent Evolution of Domains in Immunoglobulin Light Chains? A. B. Edmundson,* K. R. Ely, E. E. Abola, M. Schiffer, and N. Panagiotopoulos

ABSTRACT: The crystallographic structure of a human

(Mcg) A-type light chain (Bence-Jones) dimer has been correlated with the amino acid sequence determined by J. W. Fett and H. F. Deutsch ((1974), Biochemistry 13, 4102). Each light chain is composed of amino (“variable” or V) and carboxyl (“constant” or C) regions (“domains”), which are cylinders of &pleated sheets enclosing hydrophobic interiors. The structures of the V and C domains support the hypothesis of a common ancestral gene. A comparison of sequences of the V and C domains, with the three-dimensional structures as a basis for alignment, confirms earlier views that sequence homologies have been obscured by divergent evolution. The types of changes that have occurred are indicated by the structures of the domains. For example, the polypeptide chains of the V and C domains appear to be rotational allomers. Among the a-carbon atoms of 38 pairs of selected residues, the angles of rotation between comparable V and C constituents are 163-167O in both monomers. The rotation axes pass approximately through homologous half-cystinyl residues of the intrachain disulfide bonds. The functions of the three- and four-chain layers of antiparallel P-pleated sheets constituting the basic “immunoglobulin fold” have changed during the evolution of the two domains. I n the light chain dimer the three-chain layers

T h e serum IgGl immupoglobulin and the urinary BenceJones protein from a patient (Mcg) with multiple myeloma and amyloidosis have been crystallized and characterized by chemical and crystallographic techniques (Deutsch, 1971; Deutsch and Suzuki, 1971; Fett et al., 1973; Fett and Deutsch, 1974; Schiffer et al., 1970, 1973; Ely et al., 1973; Edmundson et al., 1970, 1971, 1973, 1974a,b). The IgGl protein consists of two light (mol wt 23,000) and two heavy chains (mol wt 50,000). The Bence-Jones protein is a dimer of A-type light chains identical in amino acid sequence with their counterparts in the parent IgGl molecule. In contrast to most IgG proteins, the Mcg molecule does not have light-heavy or heavy-heavy interchain disulfide bonds because of a deletion of 15 residues in the “hinge” region usually supplying the appropriate half-cystinyl residues (Fett et al., 1973). The light chains are of normal length and are connected by an interchain disulfide bond both in the IgGl protein and in the Bence-Jones dimer. The presence of a crystallographic twofold axis of rotation between halves of the IgGl molecule indicates that the covalently linked light chains have identical conformations when associated with heavy chains (Edmundson et al., 1970). In the Bence-Jones dimer, however, the spatial relations between the V and C domains in the two light chains are strikingly different (Schiffer et al., 1973). The angle bet From the Division of Biological and Medical Research, Argonne National Laboratory, Argonne, Illinois 60439. Received April 1 4 , 1975. This work was supported by the U S . Energy Research and Development Administration.

of the V domains face each other across a solvent channel in which hapten-like molecules can be bound. The corresponding layers in the C domains supply most of the external surfaces of the dimeric module. Conversely, the four-chain layers furnish the externzl surfaces of the v 1 - V ~dimer, but interact across a solvent-free zone to help maintain the compact arrangement of C domains. Sequence patterns in parts of the pleated sheets can be correlated with these differences. In the V domains an alternating pattern of polar and apolar residues is broken in the three-chain layers by the substitution of aromatic for aliphatic polar residues in positions important for the maintenance of the general architecture of the hapten-binding sites. Similarly, the alternating sequence pattern in the four-chain layers of the C domains is interrupted by the substitution of hydrophobic for polar residues a t sites of interactions in the C I - C ~interface. The polar constituents in the alternating sequences of the four-chain layers of the V domains are mainly seryl and threonyl residues, which form most of the surfaces of the dimeric module. The polar residues on the surfaces of the C dimer are more diversified. The differences in surface properties may partially explain why the V domains have been implicated in the formation of amyloid fibrils and in the characteristic thermal behavior of Bence-Jones proteins.

tween the long axes of cylindrical envelopes drawn around the V and C domains is approximately 70’ in monomer 1 and 110’ in monomer 2. Differences in the “switch” regions (i.e., around Gly-1 11) are mainly responsible for these conformational variants (Edmundson et al., 1974b). Comparisons of the structures of the light chain dimer and Fab fragments (Poljak et al., 1973, 1974; Amzel et al., 1974; Padlan et al., 1973; Segal et al., 1974; Padlan and Davies, 1975) indicate that the conformation of monomer 1 is similar to that of the heavy chain, while monomer 2 closely resembles the light chain components. The crystallographic analysis of the light chain dimer a t 3.5-A resolution showed that each domain consists of a basic layered structure, one layer with four antiparallel extended chain segments, and the second with three antiparallel segments (Schiffer et al., 1973). The two layers are covalently linked near the center of each domain by an intrachain disulfide bond. Similar layered structures have been found in the domains of both heavy and light chains in human and murine Fab fragments (Poljak et al., 1973, 1974; Amzel et al., 1974; Padlan et al., 1973; Segal et al., 1974) and in V domain dimers from K-type Bence-Jones proteins (Epp et al., 1974; Fehlhammer et al., 1975). Extension of the crystallographic analysis of the Mcg dimer to 2.3 A indicated that the layers in each domain are composed of “twisted” &pleated sheets (Edmundson et al., 1974a). These sheets enclose hydrophobic interiors. When examined in isolation, the structures of the V and C domains strongly support the hypothesis of a common ancestral protein of about 110 amino acid residues (Singer BIOCHEMISTRY, VOL. 14, NO. 18, 1 9 7 5

3953

E D M U N D S O N ET AL

C

FIGURE 1: Tracing of a photograph of a model of the Bence-Jones dimer, with the polypeptide chains of monomers 1 and 2 shown in white and black, respectively. The solvent channel in which hapten-like molecules can be bound lies between the V domains on the right. The amino-terminal residue in each domain is labeled N, and the interchain disulfide bond between penultimate residues is located on the extreme left. Note the difference between the spatial relations of the V and C domains in the two monomers. Monomer 1 has a conformation similar to that in heavy chains of Fab (antigen-binding) fragments, while mcnomer 2 closely resembles the light chain components. This figure was adapted from Schiffer et al. (1973).

and Doolittle, 1966; Hill et al., 1966). However, the threeand four-chain layers have markedly changed functions during the evolution of the V and C domains. In previous articles we did not attempt to define the structural basis for these functional changes. The availability of a 2.3-A electron density map (Girling et al., in preparation) and the completion of the amino acid sequence determination (Fett and Deutsch, 1974) nbw permit this problem to be considered in the present report. Materials and Methods Details of the crystallographic study leading to the calculation of a nominal 2.3-A electron density map will be published in another article. Using the amino acid sequence (Fett and Deutsch, 1974), Watson-Kendrew skeletal models were fitted to this map with the aid of an optical comparator (Richards, 1968). Atomic coordinates were measured from the model and subsequently used in Diamond’s model building programs (1966). Structural analyses, including comparisons of the V and C domains, were first performed by direct observation of the model. A least-squares procedure was then used to compare the a-carbon positions of selected residues in the V and C domains. The domains were superimposed by translation, and rotational parameters were adjusted until the sums of the squares of the distances between equivalent atoms were minimized (Huber et al., 1971; Drenth et ai., 1972; Ohlsson et al., 1974). Results and Discussion For correlation with the sequences to be discussed, a tracing of a photograph of a scale model (2 cm = 1 A) of the dimer is shown in Figure 1. Comparisons of amino acid sequences in light chain domains are difficult because of the apparent divergence of V and C genes (for reviews and sequence alignments, see Edelman and Gall, 1969; Milstein and Pink, 1970; Garver and Hilschmann, 1971; Dayhoff, 1972; Putnam, 1975; Poljak et al., 1974). With the aid of the following guidelines, however, the sequences can be compared within the framework of the three- and fourchain layers in the crystallographic structure. A schematic drawing of monomer 2, with directional arrows superim-

3954

BIOCHEMISTRY, VOL.

14, N O , 18, 1975

15

2: Schematic drawing of monomer 2, with directional arrows superimposed on segments participating in antiparallel 8-pleated sheets. Three-chain layers are indicated by numbered striated arrows and four-chain layers by white arrows. Positions of representative residues are numbered to aid in the correlation of amino acid sequences with the three-dimensional structure. FIGURE

posed to indicate the locations of the three- and four-chain layers, is presented in Figure 2. The complete amino acid sequence of the Mcg light chain in reference to the threedimensional structure is given in Figure 3. The intrachain disulfide bonds are in similar locations in the two domains, and the corresponding disulfide loops are approximately the same size (-60 residues; see above reviews). Our sequence alignments were first confined to the most regular parts of the pleated sheets, since the threedimensional structures of the V and C domains are most similar in these regions. About eight consecutive residues in each segment of the three- and four-chain layers were aligned in positions relative to that of an intrachain halfcystinyl residue. The chosen sequences, together with the patterns of emergence of side chains on alternating sides of the three- and four-chain pleated sheets, are illustrated in Figure 4. Residues selected for comparisons of a-carbon coordinates are underlined. Less ordered regions and turns were compared next. Residues 23-36, 51-55, and 91-99, statistically defined as “hypervariable” in different light chains (Wu and Kabat, 1970; Kabat and Wu, 1972), were treated with caution. The hypervariable regions have few sequence identities with corresponding segments in the C domains, and include residues involved in insertions and deletions in other light chains. Rotational Allomerism in the V and C Domains. The similarities in the cylindrical domains are illustrated in Figure 5. The schematic drawing of the C domain is rotated to approximate the orientation of the V domain. Colman et al. (1974) have evidence for the presence of approximate twofold rotation axes (with translations) between the C H and ~ C H domains ~ of the Fc (“crystalline”) fragment of an IgG immunoglobulin, and we looked for similar axes in the light chain monomers. The search was initially limited to observation of the atomic model in structurally important regions, in which sequence differences are minimal. Such segments are found near the intrachain disulfide bonds, which are shielded in each domain by a tryptophyl residue (37 and 152). The conformations of the tripeptide segments Val-Ser-Trp (35-37) and Val-AlaTrp (1 50- 152) bear an unmistakable resemblance to each other. As indicated schematically in Figure 6, these two sets of three residues are related by a noncrystallographic rotation axis, and a translation of 43-44 A in monomer 2. This axis passes approximately through the side chains of Cys-22

ROTATIONAL ALLOMERISM IN IMMUNOGLOBULIN LIGHT CHAINS 10 20 30 40 50 60 ZSALTQPPSASGSLGQSVTISCTGTSSDVGGYNYVSWYQQHAGKAPKVI IYEVNKRPSGVPDR

Q - - - 4 - 1 - - - 0 ~ Q - 4 - 2 - - 0 ~ ~ 3 3 - 1 - - - 0 ~ ~ / L ~ V DOMAIN

70 m Po 1w 110 FSGSKSGNTASLTVSGLQAEDEADYYCSSYEGSDNFVFGTGTKVTVLG

uo 130 140 150 160 170 QPKANPTVTLFPPSSEELQANKATLVCLISDFY P G A V T V A W K A D G S P V K A G V E T T K P S K Q S N N - 4 - W

-4-2-

-

-4-4-m

-3-1-

C DOMAIN

K Y A A sim s Y L s L T P E Q w 190 KsHR

s Y s C Q v200 T H E G S T V E K T210 VA PTEc

m +3-2+

+4-3---0

EXTENDED CHAIN

I HAIRPIN TURN

s-

3 +3-3--0

HLlX OTHER STRUCTURE

FIGURE 3: Amino acid sequence of the Mcg X chain, as determined by Fett and Deutsch (1974). Residues are represented by one-letter abbreviations (Dayhoff, 1972). The three-dimensional structures assumed by the segments are indicated schematically under the appropriate sequences. Extended chains participating in the three- and four-chain layers are numbered as in Figure 2. The symbol for “other structures” indicates corners other than hairpin turns, as well as irregular structures not included in other categories.

4- 1

~~~

~

*

114 115 116 117 118 119 120 121 122 li3 K

A

N

U

T

I

T

U

F

3-1

I

I

*

*

N

Y

n

S

b

Y

33

34

35

36

37

UI 39

P

: *

*

I

o

Q

H

40

41

1

147 148 149 150 151 152 153 154 155 T

n

T

m

S

n

T

B

S

Q

25

24

23

22

21

20

19

18

17

16

A

T

[II’A

K

D

8

4-2

I

* : * I

1 I

201 2M M

T

199 198 197 ,196 195 194 193 a

Q

Q

I

S

U

R

S

I I

4-3 175 176 177 178 179 & ll K

Y

A

O

S

Q *

181 182 183 180 185 Y

I

U

S

O

*

I

8

I

N

*F D*F h ’ nnKUTBt

9%

PP

T

io0 101 102 io0 io5 LW io7 io8

110

3-3



2w 205 206 207 Ea 209 210 211 S T I E T U A

I

I

4-4

im 169 168 167 166 165 164 163 162 161 K

S

I

K

* O:

T

E

D

G

A

FIGURE 4: Amino acid sequences of segments in three- and four-chain layers of @-pleatedsheets. Equivalent pairs of residues are listed, with the V components on top. Abbreviations of residues with side chains oriented toward the interior of each domain are enclosed in boxes. Residues marked with asterisks in the three-chain layer of the V domains are constituents of the sites in which hapten-like molecules are bound in the light chain dimer (see Figure 8). In the four-chain layer of the C domain the residues designated with asterisks occupy the interdomain space and help to stabilize the C I - C ~dimer (see Figure 9). The vertical dashed lines pass through positions approximately aligned with the intrachain half-cystinyl residues. There is a “kink” in chain 3-3 a t Thr-103, which consequently is not in the same relative position as Thr-209. The latter is aligned with Gly104 in the figure.

and -138. Because of the differences in the spatial relations between the V and C domains (see Figures l and 7), the rotational allomerism is easier to visualize in monomer 2 than in monomer 1 . Nevertheless, the above tripeptide segments of monomer 1 can also be related by rotation and translation. When 38 pairs of residues (underlined in Figure 4) were compared by a least-squares procedure, the average deviation between the a-carbon positions of equivalent residues was 1 . 1 1 8, in the VI and C1 domains and 1.16 8, in the V2

and CZdomains.’ The average value for the angles of rotation between equivalent a-carbon atoms in the V and C domains were 163O for monomer l and 167’ for monomer 2. These results indicate that the polypeptide backbones in the ~

’ For the same 38 pairs of residues in like domains, the correspond-

ing values were 0.94 8, for the V I and V2 domains and 0.95 8, for the (21and C2 domains. When all pairs of a-carbon atoms were compared, the angle of rotation between V I and V2 components was 180°. The C I and C2 domains were also related by a twofold axis of rotation. BIOCHEMISTRY, VOL. 14, NO. 1 8 , 1915

3955

EDMUNDSON

ET A L

(a 1 (b) 5: Comparison of the V and C domains of monomer 2. The C domain (a) is rotated to approximate the orientation of the V domain (b). The directional arrows in the three- and four-chain layers correspond to those in Figure 2. Chain segments present in only one domain or having different structures in the two domains are indicated by solid black lines. Positions of prominent residues, or their near-neighbors, are numbered for correlation with the text. FIGURE

6-pleated sheets of the V and C domains are closely similar in structure and in the characteristic twist of the chain segments. The similarities of monomers 1 and 2 to the heavy and light chain components of the Fab fragments strongly suggest that rotational allomerism is a fundamental feature of immunoglobulin structures. Alternating Polar-Apolar Sequences in the @-Pleated Sheets. The recognition of rotational allomerism in the V and C domains permits the comparison of individual residues, particularly in the &pleated sheet regions. In pleated sheets the side chains emerge at right angles to the direction of the polypeptide chain and on alternating sides of the backbone. Given a cylindrical domain with a hydrophobic center, we looked for sequence patterns with external polar residues alternating with internal apolar residues. Such patterns are readily seen in the four-chain layers of the V domains and in the three-chain layers of the C domains (see Figure 4). In the V domains the polar residues in the alternating patterns are mainly serine and threonine, and the cylindrical surfaces are largely composed of side chains of these hydroxy amino acids. The surface residues of the C domains are more diversified. The alternating sequence patterns are interrupted in important locations in the three-chain layers of the V domains and in the four-chain layers of the C domains. The significance of these interruptions will be considered after a discussion of the side chains directed toward the interior of each domain. Internal Residues in the V and C Domains. Each cylindrical domain is filled with hydrophobic side chains, which are listed in Table I (also see Figure 4). Where possible, the residues are paired in equivalent positions in the two domains. The intrachain disulfide bonds are well shielded from solvent by the side chains of Leu-4, Gln-6, Ile-20, Val-35, and Trp-37 in the V domains, and those of Val-119, Leu-121, Val-150, Trp-152, and Val-199 in the C domains. Gln-6, (3111-39, Ser-92, Thr-105, Thr-165, and Ser-180 are exceptions to the usual distribution of hydrophobic residues inside the cylinders. The observations indicate that approximately one-fourth of the residues occupy internal sites in each domain. This

3956

BIOCHEMISTRY, VOL.

14, NO. 18, 1975

F I G U R E 6: Schematic drawings showing the noncrystallographic axis of rotation between the V and C domains of monomer 2. (a) Monomer 2, as presented in Figure 2, with the rotation axis included. For a-carbon atoms of 38 pairs of residues selected from the three- and fourchain layers of the V and C domains, the angles of rotation are 163167'. (b) Same view, with the domains represented by cylindrical envelopes. After rotation, corresponding structural elements in the two domains have to be translated 43-44 8, to bring them into coincidence. (c) Illustration of the rotational symmetry between two corresponding tripeptide segments near the intrachain disulfide bonds of the two domains.

fraction is comparable with that in the myoglobins and hemoglobins (Perutz et al., 1965), and is consistent with earlier conclusions based on calculations of average hydrophobicities and on the distributions of polar and apolar residues in the V and C domains of series of light chains (Welscher, 1969a,b; Edmundson et al., 1973). In the globins, which have similar three-dimensional structures, only a few (